Good catch, thanks for your contribution!

This also seems to be the behavior for when a synchronous entry point handler returns None, although this is never explicitly documented. Perhaps it would be a good idea to do so.

It would also be great if you could add a unit test that tests for this bug, so we won't have any regressions in the future. It should probably look similar to this test.

Edit: I don't think the caveat you mentioned is real, pretty sure it is already not possible to use None for states

@vasinkd is test and docs something you would like to do an attempt at?

Totally forgot about this. @jsmnbom , thanks for a reminder!

Ok, I'll add tests, no problemo.
What about documentation? Should I mention this behavior in ConversationHandler class description?

Thanks a lot :D
In terms of docs I was mostly referring to @jh0ker's comment about

This also seems to be the behavior for when a synchronous entry point handler returns None, although this is never explicitly documented. Perhaps it would be a good idea to do so.

So a note in the ConversationHandler class docstring about this behaviour (both sync and async) sounds good yea :)

Done. Sorry for the delay!
Better late than never :)

Test for this instance passed. Can we merge for V12 @jsmnbom ?

CI is running, should be able to be merged in v12

Merged V12 after making small adjustment there regarding conversationhandler. If CI agrees we can merge.